HP Investigating Face-Tracking Bug Reports

  • Thread starter Thread starter Math Is Hard
  • Start date Start date
  • Tags Tags
    Bug
Click For Summary

Discussion Overview

The discussion revolves around the potential issues with Hewlett-Packard's motion-tracking webcams, particularly regarding their ability to recognize faces of individuals with darker skin tones. Participants explore the implications of contrast recognition in computer vision and share personal experiences related to webcam functionality.

Discussion Character

  • Debate/contested
  • Technical explanation
  • Conceptual clarification

Main Points Raised

  • Some participants highlight that HP's webcams may struggle with contrast recognition in certain lighting conditions, which could affect their ability to detect faces accurately.
  • Others argue that this issue is a common problem in computer vision, where detection algorithms often rely on contrast, making them vulnerable in low contrast situations.
  • A participant expresses skepticism about HP's solutions, suggesting that the company may have overlooked basic facial-recognition technology.
  • There is a reference to a TV show episode that humorously parallels the situation, indicating a cultural commentary on technology and race.
  • Some participants note peculiarities in the video, such as backward nametags, suggesting that there may be additional factors at play affecting the camera's performance.
  • Discussion includes personal anecdotes about webcam experiences, particularly regarding image reversal in Mac cameras, which some participants find noteworthy.

Areas of Agreement / Disagreement

Participants express a range of views on the effectiveness of HP's webcams, with no clear consensus on the underlying causes of the issues raised. Some agree on the technical challenges posed by contrast recognition, while others introduce different perspectives and experiences, leaving the discussion unresolved.

Contextual Notes

Limitations include the lack of detailed technical specifications for the webcams and the variability in personal experiences with different camera models, which may influence perceptions of the issue.

Math Is Hard
Staff Emeritus
Science Advisor
Gold Member
Messages
4,663
Reaction score
36
Can't believe I haven't seen a thread on this yet.

http://edition.cnn.com/2009/TECH/12/22/hp.webcams/

Can Hewlett-Packard's motion-tracking webcams see black people? It's a question posed on a now-viral YouTube video and the company says it's looking into it.

...

HP acknowledged in a statement e-mailed to CNN that the cameras may have issues with contrast recognition in certain lighting situations. The webcams, built into HP's new computers, are supposed to keep people's faces and bodies in proportion and centered on the screen as they move.

The YouTube video is here, and it's pretty darned funny.

 
Last edited by a moderator:
Computer science news on Phys.org
Math Is Hard said:
Can't believe I haven't seen a thread on this yet.

It should probably be in the comp sci forum. It's a great lesson on testing and robustness and all that other real world stuff that sometimes gets lost in comp sci programs.

For any non-computer scientists:
Basically, it's a standard problem in computer vision-a lot of detection algorithms rely on contrast to find features, so low contrast situations (bad lighting, similar colors, etc.) can really mess with what an algorithm finds. There are ways to correct for this, but it looks like hp just threw out a pretty basic facial-recognition solution that doesn't.
 
I have an HP and I think they threw out a many a solution that doesn't.
 
Wasn't there a "What about Ted" episode about this?
 
You mean "Better off Ted"? Yeah, there was an episode where the company installed sensors that wouldn't detect black people (so lights would turn off when they were working, drinking fountains wouldn't work for them, etc) so the company hired white people to follow around the black people...
 
ha ha
Life imitates Art!
 
Did you notice that their nametags are backward in the image? Either the camera also reverses images, or they're doing something odd like aiming the camera at a mirror (there's also a reflected glare of some other lighting in the image). It may be true that there's a contrast problem with getting a good focus, but it seems they're doing something else in that video to trick the camera.
 
Moonbear said:
Did you notice that their nametags are backward in the image? Either the camera also reverses images, or they're doing something odd like aiming the camera at a mirror (there's also a reflected glare of some other lighting in the image). It may be true that there's a contrast problem with getting a good focus, but it seems they're doing something else in that video to trick the camera.

Someone else probably knows why this is, but the cameras built into Macs automatically reverse the image as well.
 
zoobyshoe said:
Someone else probably knows why this is, but the cameras built into Macs automatically reverse the image as well.

I've never noticed that on mine. I use it often to Skype with my boyfriend, and have the PIP display on so I know I'm still in view.
 
  • #10
Moonbear said:
I've never noticed that on mine. I use it often to Skype with my boyfriend, and have the PIP display on so I know I'm still in view.
My experience was that I was drawing a portrait of a guy, which took about three weeks, and he had me hold the drawing up to his Mac several different times during the process to take work-in-progress shots of it. I saw right away that all these shots were reversed, right and left. I asked him why and he said "I dunno, the camera just does it that way." It could be it was only still shots, or, it was only certain models of Macs that had this peculiar quirk.